Research map / Label-efficient and robust learning / OOD and anomaly detection
Dual Distillation for Few-Shot Anomaly Detection
Le Dong, Qinzhong Tan, Chunlei Li, Jingliang Hu, Yilei Shi, Weisheng Dong, Xiao Xiang Zhu, Lichao Mou
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add student self-distillation on support images (ablation) | D24FAD · RSNA 2-shot | image AUROC: 57.0 -> 98.7 | Improved |
| add teacher-student distillation (ablation) | D24FAD · APTOS 2-shot | image AUROC: 92.7 -> 100.0 | Improved |
| add learn-to-weight mechanism (ablation) | D24FAD · HIS 4-shot | image AUROC: 94.5 -> 98.4 | Improved |
| add learn-to-weight mechanism (ablation) | D24FAD · LAG 8-shot | image AUROC: 98.6 -> 98.0 | Hurt |
| set loss weight lambda to 0.1 instead of 1.0 (ablation) | D24FAD · RSNA 2-shot | image AUROC: 96.7 -> 99.0 | Improved |
| use WideResNet-50 teacher instead of Swin-B (ablation) | D24FAD · HIS 2-shot | image AUROC: 64.7 -> 96.7 | Improved |
| use RadImageNet instead of ImageNet pretraining for ResNet-50 teacher (ablation) | D24FAD · Brain Tumor 8-shot | image AUROC: 96.4 -> 98.4 | Improved |
| use randomly sampled instead of fixed support images (ablation) | D24FAD · HIS 4-shot | image AUROC: 93.5 -> 93.4 | No clear effect |
| use scaled dot-product weighting instead of Gaussian weighting (ablation) | D24FAD · HIS 4-shot | image AUROC: 97.4 -> 98.4 | Improved |
| D24FAD vs MVFA-AD on Brain Tumor 8-shot | few-shot methods · Brain Tumor 8-shot | image AUROC: 96.3 -> 95.5 | Hurt |
Closest papers by meaning
- UniVAD: A Training-free Unified Model for Few-shot Visual Anomaly Detection (CVPR 2025 · similarity 0.87)
- Adapting Visual-Language Models for Generalizable Anomaly Detection in Medical Images (CVPR 2024 · similarity 0.87)
- Scale-Aware Contrastive Reverse Distillation for Unsupervised Medical Anomaly Detection (ICLR 2025 · similarity 0.87)
- Unraveling Normal Anatomy via Fluid-Driven Anomaly Randomization (CVPR 2025 · similarity 0.86)
- Beyond Normal References: Discriminative Few-Shot Anomaly Detection (NeurIPS 2026 · similarity 0.86)
- Unlocking the Potential of Reverse Distillation for Anomaly Detection (AAAI 2025 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
